Tensor-based Nonlocal MRI Reconstruction with Compressed Sensing

Compressed sensing(CS) is a significant technology in MRI reconstruction, which can reconstruct the image with few undersampled data and speed up the imaging. The conventional CS-based MRI is implemented on the global image, which not only loss many local structures but also fails in preserving the detail information. To improve the reconstruction quality, we proposed a novel CS-based reconstruction model, which is incorporated with nonlocal technology to gain extra details preservation. The proposed model grouped the similar patches within the nonlocal area, and stacked them to form a 3D array. Then, to process the array in a realistic 3D way, a tensor-based sparsity constraint is developed as the regularization on the reconstructed image. Experimental results show that the proposed method is more effectiveness and efficiency than the conventional ones.